Ep. 375: Did OpenAI Just Kill Social Media?  ## Metadata - Author: [[Deep Questions with Cal Newport]] - Full Title: Ep. 375: Did OpenAI Just Kill Social Media? - Category: #podcasts - URL: https://share.snipd.com/episode/cd44581e-5e1c-4ace-a102-61a62f49973c ## Document Notes ## Highlights - **AI Text-To-Video Bypasses Production Limits** - Sora 2 lets anyone generate striking videos by text, bypassing real-world production limits. - This threatens platforms built on human-made novelty because AI can cheaply produce endless novelty. Transcript: Cal Newport If you've spent enough time online this past month, you've probably noticed a sudden uptick at unsettling, realistic, AI-generated videos. Maybe it's body cam footage of Super Mario in his cart being pulled over by police and then yelling, I can't get another ticket before peeling out. Or a horse on skis competing in a freestyle skiing competition. Or Bob Ross wrestling Abraham Lincoln. These were all generated by OpenAI's blockbuster video generation model, Sora 2, which allows you to create videos simply by describing them in words. OpenAI, however, has gone one step further and bundled up access to this model in a new app that they call simply Sora that operates like TikTok. You can easily create videos and share them. You can view other people's creations, and there's an algorithmically curated feed just like you would see on TikTok. You can have a steady incoming feed of grade A high-octane LOP. The online world is usually pretty excited about new AI products, but the reaction to Sora has been more interesting. People online seem, to put it simply, unsettled. I want to play a clip here. Here's Casey Neistat reacting to the new Sora app. My analysis? Jesse We are so completely... Cal Newport And now here is vlogger Hank Green venting some of his own frustrations. Jesse Boy, humanity, if you're the kind of motherfucker who will create slop talk, you are not the kind of motherfucker who should be in charge of open AI. All right. Cal Newport The New York Times, for its part, was really worried about what this means for even just truth. They quoted a founder of a tech nonprofit saying the following. Nobody will be willing to accept videos as proof of anything anymore. All right. But for all of these particular fears we're hearing now about Sora, there is one fear in particular that I think is really interesting but is not getting discussed that much yet. And that is a question of what will the impact be of apps like Sora on the existing major social media platforms? Think about it. It's an app on your phone in which anyone can basically create any visual content they can think of. Why would I watch a TikTok dance routine? When using Sora, I can create a video of Michael Jackson dancing on Mount Everest with Martin Luther King Jr. That's actually a real video I found, Jesse. So by introducing Sora, has OpenAI just radically changed the entire social media landscape? This is the question we're going to get into today. ([Time 0:00:02](https://share.snipd.com/snip/14efe3bb-c36e-498a-ac7f-16694bce387d)) - **Early Platforms Had Durable Moats** - Facebook, Instagram, and Twitter became hard to displace because of their unique user bases and painstaking social graphs. - Those social-graph advantages made each platform offer distinct, hard-to-replicate value. Transcript: Cal Newport So if we really want to understand the danger that the existing social media platforms currently face, I want to take you back in time to roughly the summer of 2022. Back then, there is a new entrant in town called TikTok, and it was just taking off and having this sort of phenomenally fast user growth. And the entrenched social media platforms, they were horrified by the speed with which TikTok was growing and the fear that it was pulling their users over into its own ecosystem. So the existing social media incumbents responded the way you might expect. They tried to make their services more like TikTok. This means they began to ignore the way they classically operated with friends and followers and favorites and retweets and instead try to offer up a stream of content selected exclusively By an algorithm to be as low friction as possible and as engaging as possible. I noticed this shift with some interest back when it was happening. I wrote an article about this for the New Yorker back in the summer of 2022. It was called TikTok and the Fall of the Social Media Giants. And that article argued that this move to be more like TikTok was something that they would come to regret. So why don't I load this article on the screen here for those who are watching instead of just listening. All right, so here's the article. Pretty cool graphic. I don't know if I ever really understood it, Jesse, but it seems to be a lot of people in cages, but the doors are open and they're circular cages and they're dancing. The cool graphic, I never quite understood it. All right, so let me start with the following thing I wrote in this article. Introducing the idea that the other social media platforms, the existing platforms, were chasing TikTok, I wrote, this shift is not surprising given TikTok's phenomenal popularity, But it's also short-sighted. Now to unpack that, let's start by asking a fundamental question. What was it that made the existing social media giants so unassailable before the era of TikTok began? I want to go through three of them one by one. Let's look at them carefully. We'll start with Facebook. When you're Facebook, ([Time 0:03:18](https://share.snipd.com/snip/b43e4cb4-c0a7-42f9-9060-10a6820e149c)) - **TikTok Bypassed Social-Graph Defenses** - TikTok succeeded by removing requirements for known users or influencers and relying on algorithmic recommendation. - That bypass undermined incumbents' social-graph defenses and shifted competition to pure engagement. Transcript: Cal Newport And the thing about TikTok is that it doesn't really care about who its users are. And TikTok did not really care about who knew who, follower graphs, favorites, interests. They didn't really care about that either. They basically bypassed the elements that made the social media giants hard to compete with. I want to bring up on the screen here, this is the way I explained it in my article. So why don't, why don't I just read this? Cause I think it, it captures it well. The effectiveness of the TikTok's experience is found in what it doesn't require. Unlike Twitter, TikTok doesn't need a critical mass of famous or influential people to use it for its content to prove engaging. The short video format grabs users' attention at a more primal level, relying on visual novelty or a clever interplay of music and action or direct emotional expression to generate Its appeal. And unlike Facebook, TikTok doesn't require that your friends already use the service for you to find it useful. Though there are some social features built into TikTok, they're not the main draw of the app. TikTok also doesn't rely on its users to manually share content with friends or followers to surface compelling offerings. It assigns this responsibility to its scary good recommendation algorithm. All right, so TikTok bypassed the competitive advantages that protected the social media giants by just saying, look, we're just going to take a whole bunch of content. We don't care who creates it. And we'll use a very good, probably like two tower style recommendation app to just show you a constant stream of stuff. We'll think you like it. We'll adjust what you show you based on your behavior. ([Time 0:10:56](https://share.snipd.com/snip/aa4d19a7-cd33-4108-8ce6-3e265e6ab96a)) - **Pure Engagement Is A Losing Long-Term Game** - Pure engagement becomes a commodity where platforms compete with all other attention sources. - Losing a unique value proposition means being vulnerable to any new, more engaging distraction. Transcript: Cal Newport So basically what I'm saying here is all of social media was collapsing around the TikTok model, which was just pure engagement. And my argument here is if all you are offering to people is pure engagement, you are now competing with every other possible source of engagement that exists out there. When Facebook is saying your friends are here, you can see what your friends are up to. I'm not competing with Netflix. I'm not competing with podcasts or slot machines or pornography or anything else because I'm offering you a very specific thing that only I can do. But when I'm instead just giving you like a long, infinite feed of videos from people you don't know or really about that are just trying to capture your attention, then that's just purified Engagement. And you're competing with anything else in that moment that might grab my attention. And so I ended that warning by saying, new things will come along that are even more engaging than you. You don't have a competitive advantage. You can't stay on the pedestal for much longer. All right, this brings us back to the main argument for today. We can think of the Sora app as one of the first major sort of new distractions or entertainments that I warned about in that article. Once all these social media platforms said, we're just going to tickle your brainstem, OpenAI came along and said, well, we can do that even better. But you have to actually get people to create these videos and they're confined by their environment and the people they have the film and they're not professional actors. They don't have good lighting. And like really interesting stuff is dangerous and hard to do. And only like a few number of video creators actually have enough money to do that. Like we can just do that all in AI. We can make if all you're looking at is like, I don't care what this video is. I'm just swiping, swiping, swiping. ([Time 0:13:40](https://share.snipd.com/snip/482efa9e-4909-429d-a5af-552522f91e0f)) - **Refocus Twitter On Real People And Curation** - Twitter should abandon autoplay TikTok tropes and refocus on hosting interesting people and distributed curation. - Stabilize content moderation and invest in features that reduce friction for real-time human conversation. Transcript: Cal Newport I'll say, Jesse, as an aside, it is a little bit weird that me, Mr. Quit Social Media, is trying to help these companies out. But honestly, I don't think there can be like kind of a worse app than what we see with Aura, Sora. So to me, this is kind of like an enemy of my enemy is my friend type of situation. I mean, I guess I wouldn't be that unhappy if all these services went out of business, but, you know, I would like them to be better than Sora. So I'm going to do this. I'm going over the dark side and I'm going to give advice to the major social media companies. All right, let's start with Twitter. Here's my advice for Twitter to fight off the AI threat. One, forget trying to do that video autoplay TikTok type thing that you've been doing where, you know, someone clicks on a video that they really want to see, and then you try to get them Into an algorithmically curated feed. That is the exact type of thing that makes you vulnerable to Sora. Instead, you have to work, I would say, on the competitive advantage that always made you strong. Be a place where interesting people go to say interesting things and that you have this like really good distributed curation model of capturing like what is interesting to the zeitgeist On any given day. To do this, I would say Twitter probably has to find a better middle ground on content moderation. Right before Elon Musk took over, the service was becoming like a lot of institutions to be holding to the far left. And this was having a sort of stultifying effect, and it was really also sort of like annoying and kind of censorious. After Musk has gone too far in the other direction, so he kind of went too far the other way. You scroll Twitter for five minutes, you get the Jew hatred, which to me is always like the of like, okay, this is this things aren't going great here. We also have a lot of arbitrariness from Musk himself where he'll just sort of like someone or not like someone and put his finger on the thumb. None of that is good if you're trying to create a town square that you want a lot of people to go on to to see what's interesting in the day. So find some sort of middle ground, like build a council. We have a council who's in charge of all content moderation policies. And by definition, it's going to be two people left-leaning, two people right-leaning, and like two centrist. And they vote on all the policies, and it's transparent, and we apply them clearly. And just be reasonable. People are okay with reasonable. If you do that and then keep focusing on feature improvements, making the experience better, like get the friction out of here, make it easier to do this or that. I think you could still do well in that model. I think Twitter can be a very sustainable three to four billion dollar annual revenue business and still have an impact on the culture. Yeah, you're not going to be meta. That's pretty good for a private company. And you're pretty AI proof because this is all about like real people saying things that are interesting in the moment that are being selected in a way that's distributed intelligence In a way that like an algorithm can't quite replicate. ([Time 0:21:37](https://share.snipd.com/snip/40cea3af-a246-4d10-a519-48bf2f503e0d)) - **Return Instagram To Creator Discovery** - Instagram should return to promoting influencers and creator follow relationships rather than pure algorithmic reels. - Prioritize discovery of creators to rebuild a stable, interest-based feed you select. Transcript: Cal Newport I think Instagram needs to return to its bread and butter of interesting influencers producing visually interesting content and allowing users to specify, oh, I like this person. I want to see what they're doing. So moving, retreating from their push towards algorithmically curated sort of reel after reel of just like, we'll just show you stuff that might be interesting. What the algorithm on Instagram should refocus on doing is just suggesting new people that you might want to follow or look at. Hey, you like these type of authors, you might like this one and you go check them out. And you're like, yeah, I like what they're doing. And you add them to the stable of people that you're listening to. But again, Instagram should be a place where they're sort of like experts doing stuff you find interesting, producing content that's visually interesting. And you have a good feed of that from people that you selected. That again is going to differentiate you from a sort of the slop style world that you'll see something like on something on Sora, because it's not just random engagement. It's engaging content on specific things you find interesting. It builds on like human expertise and human connections. AI approves you. ([Time 0:24:22](https://share.snipd.com/snip/4c8db95c-7867-45c2-8edf-f08104cc7b39)) - **Reclaim Facebook's Friend-Graph Value** - Facebook should re-emphasize the friend graph and attention reciprocity: show what people you know are doing. - Lean into supporting meaningful attention exchanges that AI cannot replicate. Transcript: Cal Newport What about Facebook? You got to go back to the friend graph. That's my argument there. That Facebook should be a place where you go where you want to see what people that you know or organizations that you know about and specifically chose to follow. What are they up to? And maybe what it is that they're consuming. That like one degree of direct indirection content sharing is fine. Here is articles that my friends are reading or suggesting right now. That's fine. Or I follow a magazine. And so in my sort of news feed will be various things that they publish that, that sort of in between era of Facebook, where it was still very friend and follower graph, friend graph focused, But you had a news feed that old from the different people you knew, but not going so far that like, we're just going to make this feed all the way algorithmic and just try to like keep you Engaged. That's a really good competitive advantage. The other thing you should lean into if you're Facebook is attention to the users. This was Facebook's original business model. I wrote about this in my book, Deep Work. We forget this now in our current age of engagement in TikTok. But the original pitch for Facebook was not just your friends are here, but that they will pay attention to what you're doing. Before Facebook came out, we had had this Web 2.0 revolution where we had things like blogs that made it easier to publish content online. You didn't have to directly edit HTML files. You could, like, type into a text box and click publish. We take that for granted, but it was a big deal in the earlier 2000s. The problem of the early Web 2 era is that when people tried publishing their own things and starting their own, like, blogs on blogger.com, no one came. Because the vast, vast majority of what people have to say or write is really pretty boring. And that was really painful. And then Facebook came along and said, hey, here's a place where you can publish stuff and people pay attention because we have this social compact. If you friend these people on Facebook, you will like or comment on the stuff they do and they'll like and comment on the stuff you do. That is a really powerful feeling, that feeling of I put something out there and some people reacted to it. Professional writers like myself sort of take this for granted, but it was a really powerful pull that made people really like Facebook, especially when they moved the mobile and put The like button on there. That really got their engagement numbers up because people loved it. People love seeing other people paying attention to stuff they do. Go back to that model, Facebook. That's a good model. The people you know and the specific organizations you like are here. You can see what they're up to and what they're publishing. And you can get some attention for the stuff you put up there. Put up an article you find, then a few other people you know will comment on it. That is really powerful stuff. And it's stuff that you can't replicate with AI. The Sora TikTok experience is still a very different thing. That's just pure engagement. This is something that's going to be hard to do if I don't have all the users on there and all of the friend graphs. So that's what I would do if I was Facebook. ([Time 0:25:28](https://share.snipd.com/snip/0982e954-039f-4a2c-a3cd-b2973161c801)) - **AI Video Has High Creation Costs Today** - Sora-generated video creation is computationally expensive, so supply and quality are constrained by cost. - That cost advantage preserves incumbents like TikTok for now because user-generated video is cheap to produce. Transcript: Cal Newport It is expensive to produce Sora videos using the Sora 2 video generation model. It is computationally very expensive. You have expensive GPU chips that are fired up for a long amount of time. It's a huge amount of electricity and a huge amount of usage to produce that 30 seconds of Bob Ross, you know, wrestling with Abraham Lincoln. So the only way for this to be economically feasible, of course, is that OpenAI has to charge for the creation of these videos. So right now you can download the Sora app for free. But if you want to create a video, you have to have at the very least a ChatGPT Plus membership. This is at least my understanding, which is $20 a month. If you have that $20 a month membership, you can produce up to 50 videos a month, but they have to be at 480p resolution, which is low. And it's actually gonna be kind of hard to compete with existing videos that are out there. And 50 might sound like a lot, but you also have to keep in mind the way these videos are created is you just type in some words and hit generate. You can imagine it might take a lot of iterations on your descriptions to get just one video that you like. But if you really want to be producing a reasonable number of Sora videos that you can iterate on at a reasonable resolution that could actually do good on a lot of these high-resolution Devices, you have to have the ChatGPT subscription account that's $200 a month. That's a really high price point when you compare it to TikTok's price point, which is free for all users because the cost of operation here is so much less. ([Time 0:29:07](https://share.snipd.com/snip/31137b29-8347-4424-8845-27f684b4da94)) - **Recommendation Engines Need Inventory** - Recommendation quality depends on inventory; Sora's costly generation limits inventory and matching. - Scarcity of AI video inventory weakens its ability to replace platforms with massive user-generated content. Transcript: Cal Newport So when you're doing recommendation algorithms, it's kind of like a core idea. You need inventory. The more things you have that potentially suggest to someone, the closer you can become to maximizing the expected reward for the individual users and the better the experience comes So that's going to be the Achilles heel the Achilles heel of Sora potentially is that it's so expensive there's just not enough inventory for it to have the same sort of like wow you're Really matching my interest tightly that like TikTok is able to get and so what's going to happen is the relatively small number of people creating the Sora videos are just going to jailbreak Them out of the app and upload them on TikTok anyways, where they can just be part of the big inventory that TikTok is using. So TikTok might survive SOAR because I don't think the economic model necessarily makes sense. But they're still in trouble because their model, this is like a horde of barbarians is coming at their Roman gates. They might fight this one off, but there's going to be wave after wave after wave. When your model is just pure engagement, there is going to be so many other things that come along that offer that engagement better. Maybe you fight off this one, but like at some point there's going to be another app where it's going to be like sports betting with topless women that like sends you random, you know, Jackpot, uh, plot machine rewards or something while misting you with fentanyl mist. I mean, you're never going to win this game forever. The engagement, there always will be new things to figure out how to be more engaged. ([Time 0:32:02](https://share.snipd.com/snip/4863caed-9609-41ec-b5e4-eb1a38533cb0)) - **Use AI To Help—But Know The Code** - Use AI coding tools to speed work but avoid 'vibe coding' when stakes exist; verify and understand generated code. - Learn enough of the language to audit, debug, and safely integrate AI outputs. Transcript: Jesse Would my time be better spent writing code from scratch or am I better doing the best thing by focusing on the problems and using all the tools that I have available to make solutions quickly? Can I tell you what was programmed from scratch? No AI tools, your Halloween lights, Halloween lights demonstrate. Cal Newport They were timing down to the millisecond level, all built within a central timing loop elegantly. So at each iteration of the loop, like you're just in the iteration of a loop and you have to keep track of what lights are on and not on. And how do I to move that and where is it i got to randomize this thing and have an explosion have you just always known how to do that stuff or i yeah i've known how to like that type of coding Like so i'm not like a full stack engineer anymore like i don't really know like the latest systems if you tell me to build like a commercial piece of software i was like i've been out i'm A theorist i don't know but i've been coding since i was like seven, but like that type of thing, give me a blank slate. Like here's a microcontroller, like build a whatever. That was like 20 minutes. I sit in the office. It's like, let's just roll, you know? And I went for it. I actually had a job for a while in college where, uh, there was a product. There's a company in my town and they built a product. It was, it was not that interesting, but it was like a piece of um it was for measuring the optical properties of film and it would like rotate things and shoot lasers through prisms and Make measurements or whatever and the whole thing was run from a computer and they had like a special control card you would stick onto the the motherboard of the computer that would Allow you to like precisely control this thing but then the technology for using those old control boards is getting outdated. And so they're like, we, what we really need to do is like control this whole thing by a microcontroller that's on the device itself. And the computer should just like send it a command, like, Oh, do a measurement. And then like the microcontroller can like precisely, you know, control everything and then just send the data back. And my job was to do all that. Like, here's the microcontroller, here's the thing and i programmed it from scratch right so like i got really good and that was an assembler nowadays like my halloween thing i can program It's like a c or c plus plus variant jar do we know stuff but back then it was assembler like you were like you were in it and you were counting the micro uh the number of the number of clock Cycles required per instruction you could count those up to get timing. Like, okay, each clock cycle is this many microseconds and the ad is going to take three clock cycles. But yeah, that type of stuff I love. I love low level. That type of thing I can do, even though I don't build systems as a computer scientist. Anyways, what I'm trying to say, Dave, is you're not a real man if you're not programming a Simbler from scratch. Okay, here's what I have to say about that, actually more seriously about it. If you're in an engineering position and you're building a system that like there is some stakes, it doesn't even have to be something that other people are using. I'm not talking about like you're building a product that you're trying to release, but like you're building a system that does things that's going to affect other people it's helping You like update password files or review like what's going on or clean out tickets or this or that you don't want to be purely vibe coding by purely vibe coding meaning like you're just Asking ai model to produce code and your only way of measuring it is just you run it and see if it works or not and if it's not you like tell it to fix some things you need to understand a code That's running whatever these scripts are that might affect other people. You can use AI tools to help you produce that code or to help you debug that code or to give you like, hey, can you show me how I would do this thing? But you have to know enough about whatever language and system you're using to be able to verify, look at that code and see what it's doing. To be able to audit yourself for a simple script and be like okay oh i see i didn't know that command ai helped me get that so you can't be vibe coding if there's any stakes like this where It could affect like outcomes for other people you got to know what's going on it's fine i'm happy if you want to use ai to speed up the process of generating or checking or debugging code But you got to actually understand at least that's my computer science ethic, you have to know enough about that programming language to know what's going on. So AI should help you speed up something you already know how to do in this context, not do something for you that you don't know how to do. ([Time 0:51:43](https://share.snipd.com/snip/c6186d04-e718-4b5f-8710-fa2cf2a70c73)) - **Use Substack But Protect Webs Of Trust** - Substack is a strong indie-media tool; use it but resist letting it become algorithmic feed-driven. - Prefer distributed webs of trust and manual discovery over platform recommendation for newsletters. Transcript: Cal Newport It's taking advantage of the internet for distribution, the spread more ideas than would otherwise be spreadable. It's controlled by the person who writes it. It's your content. It's long form. It's idiosyncratic. So I'm a big fan of newsletters. Substack, look, there's some issues with it that catch my attention. They want to be more like a social platform. They want you to move out of your inbox into their app. They want it to be like all of these newsletters, like in TikTok, are somehow just nuggets of content. And don't worry, we'll help you find them. We'll see what you like reading and we'll find other things. You'll just kind of have this like stream of things you might be reading. I think their model is they want to be, their aspiration is to be like an algorithmically curated magazine. And I don't love that. I think email newsletters I think of as I have a reason why I want to hear what this person is writing, right? Because I'm a big believer in distributed curation through webs of trust. You don't have a machine just tell you, hey, here's something you might find interesting. You actually find your way to someone through real people. This person I already know and have a relationship with. We have a link in the web of trust. I know this person. I trust this person either because I know him personally or just through their reputation. They're now recommending this person. This person is a good writer. I can now add a link in that web of trust to this other person. Because if this person trusts them and I trust them, then I trust this other person as well. So maybe I'll check out what they're doing. I'm willing to possibly consume what they're doing. And then maybe they're constantly talking about some other writer. Well, that puts another link in my web of trust. That's now someone that through this chain of trust I can trust is like a legitimate good actor, good faith. It's interesting. It's stuff worth listening to. That's the way we used to find writers, the way we used to find books, just the way we used to find like albums and stuff like that. And I think that's a fantastic mode of curation. The internet is good for distribution in this model. I can now see what this person has to read without having to like send away for something or go find a book in store. That's fantastic. But I don't necessarily need the Internet to help me find the person. Distributed webs of trust are an incredibly effective tool, in my opinion, for curation. It handles like so many issues that we actually have about the Internet today about, you know, actors that are acting in bad faith or they're actively trying to deceive you or they're Like undermining values in ways or they're like actually like agents of like another power, this or that. A lot of that goes away when you use webs of trust because when you're building up these chains of trust, it's very hard to build your linkage over to nonsense. And it's very different, for example, than when you have a fully homogenized, algorithmically curated type of setup like TikTok. Every TikTok video is the same. Everyone's treated the same. I don't know where this video is coming from. Is this really just like someone over in, you know, like Arkansas who has an opinion on this world event? Or is this like a bot that's coming out of like, you know, Pakistan? I don't know. Or like Twitter, everything looks the exact same and stuff just kind of spreads to me that I see or don't see. I like a world where I had to go from person to person to person, making a real chain and connection each way to get to someone else. Cause I'm not going to get the nonsense that way. So I don't like that. Substack is trying to move away from that. I guess it's an engagement model. There's just a constant stream of people. You might want to hear writing that being said. So put that aside. It is a boon for writers. Substack is a boon for writers because they take out a lot of the complexity and expense of taking advantage of that distribution model. And I don't want to undercut that. I think that is actually like a really, really big contribution to the world of ideas. I do want to give Substack credit for that. I'm independent fully. I'm independent with my newsletter. That's not easy. Not easy. It's also expensive. I don't know what we pay Jesse, but I think it's like, you know, we have a hundred something thousand subscribers. It's $30,000 a year easy just to host that newsletter. Right. So it's not, these are non-trivial expenses. Now I make money because I also have a podcast that's, that is synergistic with the newsletter. And we work with a really good ad agency and we've been at this for a while. And then we have an income stream through ads that subsidize the other things. That's a really hard thing we do. And if I was just starting from scratch, my newsletter is getting popular. I don't have that type of money. Right. So I think that's a really good thing to do. It's also complicated. The software is complicated. It's not too hard. We use a really good provider, but, you know, you got to learn how this works. And Substack makes it easy. Substack also has recommendations. Like they'll, they'll recommend, as I talked about this thing that I'm a little bit wary about is like algorithmic recommendations. It does help people's lists grow, though. I'm kind of a believer in just like old fashioned list growth. It takes a lot longer, but the people you gain are like bigger on your list. So I won't cast shade on Substack because it allows writers to just focus on writing and not this like expense and complexity. I mean, like we have someone who basically just like handles the newsletter from a technological perspective. You know, Substack writers don't need to worry about that. So yeah, use it. That's fine. I think it's fine. I think it's fine to use Substack. Just be wary when you consume about, like, the algorithmic curation and, in general, use manual chains of trust to find newsletters. But I'm fine with Substack. I like that they make it cheap for people to actually do what they do. ([Time 0:56:35](https://share.snipd.com/snip/5313cf53-be2c-41c5-815b-0a57aefc49f1)) - **Replace Email With Clear Low-Friction Protocols** - To stop email back-and-forth, build team buy-in for communication protocols or expose a simple external interface. - Keep interfaces low-friction and hold the line on out-of-protocol messages to prevent context switching. Transcript: Jesse I think it's realistic for most people to give up their smartphone. However, if something is available only through a mobile app, what are your thoughts on a family smartphone? It's not tied to anybody in particular, and it's there to purely cover specific needs. Cal Newport I think it's a great idea. I think a family smartphone app is great. You have those type of apps on it you need for just like operating logistically in various parts of life and you grab it when you need it. Get a cheap phone. Keep it forever. Keep it charged in the kitchen. I actually just heard an interview with Warner Herzog on Conan O'Brien's podcast that he was actually talking about this. Warner Herzog like famously is one of these people like Quentin Tarantino or like Aziz Ansari who do not use smartphones or he's never really used a smartphone. He's too busy making like 30 movies a year. But he told Conan he had to buy one recently. Why? Because the parking lots in a lot of European cities need a smartphone app for you to use it. You have to use the app to actually like register your car so it doesn't get towed so he has some smartphone that he barely knows how to use and all it has on it is a parking app i think that's Great and so if you're uh if you're worried about you know having a fully serviced smartphone as being an unavoidable distraction then yeah have a family smartphone for when you need It that is a great idea patricia i uh i recognize it all right jesse we have our bonus question here right it's on a topic outside of today's episode but one that i like to talk about. Yes, we do. All right. Who we got? It's from Tim. Jesse We write monthly articles for our company's internal weekly newsletter. To minimize back and forth email communications, I set up a clear system and structure. This past month, a member didn't use the structure at all, but kept moving the conversation back to emails. Even when I tried to move things into the system multiple times. I do. This is like pure a world without email. Cal Newport If you haven't read that book, oh man, I really get into like what happens with email, why it's destroyed work and what it looks like to not be beholden the email. So Tim, I haven't talked about that book in a while, so you gave me an excuse to do so. I have three different things to mention and we can decide which of these three things are relevant to you. All right so when it comes to putting in place systems or protocols that are an alternative that just send the emails back and forth you have to differentiate between a new protocol that A team is going to use like, versus a protocol that you're exposing to the outside world, like to other teams, or internal changes. So when you say like, oh, we do an internal weekly newsletter, if there's like a team of writers who work on this together, and that's what you're talking about internally, email alternatives Have to be bottom up with buy-in. If you just come in and say, technically, I'm in charge of this team. Here is how we are communicating now. It creates resentment and friction, and it really rarely works. People just don't like it. They just don't. Even if it is like on paper, a better idea, they are going to have like a Cal Newport dartboard up that they're going to be, you know, throwing those darts at, and they're going to be aiming For the groin, if you know what I mean, right? It's not good. You got to have buy it. You have to be like, hey, let us talk about what's going on. Here's why I'm worried about us just doing all this with email. Here's the problems. Like I read this Cal Newport book about a world without email, like all this context shifting. Like I think this is making us miserable. It's making us less productive. We need some sort of protocol that's maybe a little bit more structured that will reduce the main thing that you want to reduce, which is not messages, but it's unscheduled messages To require responses because you're trying to prevent having to switch your context as much as possible. All right, what should we do? And as a team, you come up with this together. You have buy-in and you say, great, what's our check-in rhythm? Once a month? All right, we'll check back in and again a month. This isn't going to completely work. We'll fix the parts that aren't. We'll try new things. We'll try to see what parts are working. That's the way you change email with protocols within a team. If it's external, so if what you're talking about here, for example, is you run the internal editorial team, but then people from all over the company can send you stuff, like be like, Hey, I want you to put this in the newsletter or something like that. Then it's okay to have – expose a protocol interface and just say this is what we use. I think it's a really good idea. I talk about in the book, World Without Email, the way that like this is the – this is how they organize the Apollo program because there are so many different teams between academia and Through the government and through defense contractors they eventually essentially had to publish like communication protocols for communicating between the teams here is where The information comes and what format we sign it off and when you do it because otherwise it was too impossible we'd like too much information or whatever like that so i think that's fine If you're exposing a communication interface to the outside world or outside your team then you can just say this is the way we communicate. And there, I think it is often sufficient to just pretend like stuff that's not in that protocol doesn't exist. Or just say, you know, you send back, here's the instructions for like submitting to the newsletter. And like, oh, okay, but just, hey, what about this? I know what's going on here. These are the instructions for submitting to the newsletter. You kind of, that's just it. Like you just sort of hold the line on that. And a few people are like, and then they'll kind of like stop participating. Most people do it. Third thing I want to say, if you do have one of these external interfaces and you're getting a lot of this pushback, you also have to keep in mind that maybe your interface is bad. And often the issue is you have too much friction. People are busy. You're not their most important thing. The thing that they're communicating to you with through your interface is probably something that's low down on their to-do list. They just want to get it off there. You do have to respect that about people. But there's common friction points that really can make these external facing interfaces fail. And you want to make sure you're not suffering from one of these. The biggest one is like making people have to switch systems. That's a really big one. It seems simple on paper, but it really is a problem for people. If you're like, look, you got to, you're gonna have to like log in and have a login and log into some system and click on some dropdown menus. And then there's some ambiguity about like what you're trying to do doesn't quite fit into these forms. And if it's too much, like I'm making you do a lot of my work for me, you get a lot of this like internal bureaucracies. Like you got to put this in and the code here and then click from this dropdown, like what the things are like that type of friction. Like I have to switch to another system or I have to fit what I'm doing into like an interface where what I'm doing might not cleanly fit. That really annoys people. So what you want to look for is like low friction interfaces. Let me give you an example. At our department, for example, at Georgetown, like we have a weekly newsletter, you know, for the department that goes out and people come up with like things they might want to submit Or this or that. There's like a really nice protocol we have in place. And by me, it was really my depart... ([Time 1:06:38](https://share.snipd.com/snip/973670d8-7d5c-4034-8756-d2255404a91b)) - **Give Kids Shared Family Phones, Not Ownership** - For kids, treat family phones as shared logistics tools, not owned personal devices. - Limit group-chat access to supervised times and keep devices communal to avoid constant digital sociality. Transcript: Cal Newport So I think that was kind of spectacular, kind of your advice in action. My daughter now is 13, and I gave her the same dumb phone. However, she is sort of using the text threads sort of as a de facto social media. So she's on two or three different text threads. She'll wake up and there'll be 7,487 messages. And I don't quite know how to regulate that or how to help her make different choices. I'm wondering if you have any advice about that. Thank you so much for what you've been doing. And I do love the newsletter. That's a new thing for me. All right, Antonio, thanks for that question. I have strong thoughts about all this. I'm actually giving a talk. The day you're hearing this, I'm giving another talk about my kid's school, about like phones. And actually the title of the talk I'm giving today, the day you're hearing this, is when should I give my kid a smartphone? I'm giving it to elementary school age parents. So I think a lot about these things. Okay, so first of all, I like the story about your son. That's partially why 16 is considered like a good time to wait till social media because you're done with puberty. You have well-established social groups and your social identity is in place. So you're a lot less malleable at that point. And then when you bring it in, it's sort of more like when an adult gets into it. You can still end up using it too much, but it doesn't have that same impact that if you bring those tools into your life, when you're still forming your social identity, when you're still Prepubescent or going through puberty and you have all those chemicals and reconfigurations of your brain, and you're still trying to find like, you know, cleave from the parental Community, tribal diet, and sort of prove yourself into a tribe of your own. Like all these things happen around that time where you really don't want that technology, you know, in your face. That being said, when it comes to social digital sociality, there is a big gender divide. And you're seeing that, you know, these are all just bell curves and there's people on either ends of these bell curves. But in general, young women and girls are more cued in onto the social technologies than young men or boys are. So this is why your son probably had an easier time with it. You need to be wary about digital sociality. That's the terminology for sort of any socializing that's happening largely through the digital exchange of text or images. So what you would do on text threads or WhatsApp or what you would do on something like Snapchat. You got to be wary about it for a few reasons. One is when you're communicating, especially linguistically, so through text, there's a lot of the parts of your brain that are part of normal social interaction that we evolved over A couple hundred thousand years. Guardrails. Guardrails to keep you a reasonable human being and to maintain relationships in a way that's going to allow you to have a long-term relationship in your tribe. A lot of those guardrails are not activated when the communication is happening with words because they don't know that you're communicating with someone. Yeah, your frontal cortex does like rationally. You have some new nerve bundles that recognize like this text is texting to someone else, but like these deep type of social guardrails, they evolved in a time where there was no written Language. They don't know about texting, but they're tuned down. So people get nasty. People say things they wouldn't say before. With girls, it's less, I want to understand talking when I give these talks, middle school girls will tell me it's often less about like, I'm going to say something mean to you. It's way more subtle knife digging than that. It's like, okay, here's what I'm going to do. You weren't invited to this thing. And so I'm going to find a way to mention it in like a very happy way. Like, isn't that really funny or lol and put a lot of emojis or whatever. But really the whole point of that was to make you feel bad. You probably wouldn't have done that in person, right? So, so guardrails are down. It's, it's a, so it's more sharp, right? You have more opportunities for these sort of negative interactions. Two, it's exhausting. It uses a lot of mental energy to, to be in the middle of a social interaction. It's like when you come back from a party, you're exhausted or whatever, and that's fine. It should be mentally expensive, especially when you really care about it. Like you're, you know, you're people messing like teenager or whatever. But when you have those conversations and ubiquitous access to a phone, you never get a break from it. You wake up to them. When you go to bed, you're still working on them. When you're at home, you never get a break where I am just safe from having to navigate this complicated, overly sharp guardrail turned down sociology. I never get a break from it. It follows me everywhere. And that is really exhausting. And it is a burden that you don't want to place upon your kids. So digital sociality is something that we really worry about. So if your kid is using the dumb phone to basically be in a world of sort of constant digital sociality, I would change that. And there's a couple of things you could do there. The big one I would do is I do not believe in the model of here is your dumb phone. No, you don't have a phone. I have a phone. I bought a phone. I'm paying for this by the month. The family owns this phone. If you're going somewhere where for logistical reasons, you need to be able to like call us to pick you up or whatever, you can take one of the family phones with you. That's what we do in our family. We have a couple of different dumb phones. They're both terrible to use. They're not owned by any of our kids. They are there if like, okay, my oldest needs to take the city bus to baseball practice. Take one of the family phones with you today so that you can let us know if there's like a problem with the bus ride. When you get home, you put it back. So he doesn't have a sense of ownership of a communication technology. I think that is really important when you have excessive digital sociality. Then you might say, well, they can't be cut off completely. There's all this texting going on. Well, they could be cut off a lot. And what you can offer them is we will set up a family iPad that has an iMessage on it. And you can be a part of groups on that. That's fine. But your time to do that is going to be like your time to watch TV. There will be like certain times where like you can go do some group messaging now if you want. So you could still sort of be involved in these things. Half hour here, like after dinner or before dinner. But then that and downstairs where we can see you, you don't get an iPad in your room. I don't get this where people are like, I would never let my kid have a phone. But man, they're up in their room a lot, you know, with the door locked with their iPad. The iPad is just a bad phone, guys. Like you have access to all the same stuff. So you can go group text in the living room now. And if you want to do that instead of watching a show for 20 minutes, and then this iPad comes back to me, not yours. You don't own it. That is giving relief to your kid. You're like, okay, there's no way I could be communicating right now. I don't have to worry about it. And what happens is, is that their friends adapt to it. Oh, okay. So-and has like a dad that listens to Cal Newport. They all know and curse my name. Like my name kind of gets out there more and more. I'm kind of like a boogeyman, right? Or it's sort of like if in the mindset of like the American teenager, the Krampus, like evil figure that comes in is John Hite. And I'm kind of like the, he's like evil sidekick me like this less known, but sort of, I come in and help him like steal your toys. That's kind of the way they see us... ([Time 1:20:10](https://share.snipd.com/snip/d87ad937-6cd1-47b9-b259-adda97352d0e)) - **OpenAI's Product Moves Signal Strategic Stress** - OpenAI's pivot to Sora and more sensational product choices suggests financial pressure and strategic drift. - Promised AGI breakthroughs didn't materialize fast enough, pushing the company toward attention-driven monetization. Transcript: Cal Newport This was from last week. Here's Sam Altman. He starts by saying we made chat GPT pretty restrictive to make sure that we were being careful with mental health issues. We realized this made it less enjoy useful enjoyable to many users who had no mental health problems but given the series of the issue when to get this right now that we have been able to Mitigate the serious mental health issues and have new tools we are going to be able to safely relax the restrictions in most cases in a few weeks we plan to put out a new version of chat GPT that allows people to have a personality that behaves more like what people liked about 4.0 if you want your chat GPT to respond in a very human way or It starts by saying, use a ton of Emoji or act like a friend chat gpt should do it too in december as in december as we roll out age gating more fully and as part of our treat adult users like adult principles we will allow Even more like erotica for verified adults. The bad news is he goes on to clarify all of the AI generated erotica will feature him. So I don't, that's going to put his face on everyone. So that's, it's kind of like there's good and there's bad, but I guess if you own the company, a lot of, all right, let me tell you, let me tell you if you're a, an open AI monitor, what this All means. Things aren't going well over at open AI. That is the only way I can think to interpret Sora followed by this announcement of we're going to turn on erotica in chat GPT. Let me explain why. Wasn't that long ago that if you listen to Sam Altman or you listen to Dario Amadei, we were still talking in terms of the world was about to be transformed. It was, what, four months ago that Sam Altman was on Theo Vaughn's podcast comparing himself to Oppenheimer and the pending release of GP5 to the Trinity atomic bomb test. It was July that Dario Amadei was going around saying like, look, we're going to like automate half of white collar jobs. It's going to be a bloodbath. Like, you know, I don't even know who's going to be employed anymore. And we had studies coming out. They were trying to identify like who would not be affected by ai there's literally a study that i read that very helpfully identified that the safest job in the near future is going to Be a dredge operator so i guess if you're on a dredge barge operating it you're like less likely to be whatever so we were in this place until very recently where the reason why there are So many hundreds of billions of dollars being invested in these technologies is because they were going to automate huge parts of the economy. Remember Sam Altman in 2021 or 2022 put out that essay, Moore's Law for Everything, where he said this technology is going to take over everything. We have to just have a new tax system in place that basically like tax taxes, equity, wealth, because there's going to be just like three companies that do everything with their software. And you're just going to need to come and we're going to have all the money. So you're just going to have like each year, take 20% of the money and distribute it to people. So there's not riots in the street. Like that was the path they were on. Now, what are we seeing? They're saying, maybe we can sell ads against AI generated videos of Stephen Hawking being pushed into a pool. Maybe we can sell ads like among ChatGPT feeds of erotica. The change the world impact that they were on a get there or die type of trajectory that's not going to happen because gpt5 wasn't a massive leap over gpt4 it wasn't a massive leap over 4.5 they couldn't scale their way to artificial general intelligence most of their improvements on the tech side for example was coming from fine-tuning what was basically like a 4-4 model to do well on very particular tasks or benchmarks that happened to be well-suited to this synthetic dataset tuning. It wasn't becoming like generally aware. They were seeing that even though this was supposed to be 2025 was supposed to be the year of the agents, that trying to run automated processes off of LLM queries just doesn't work. They're just not reliable enough. They make things up. They don't understand what you're saying. And if you just automatically like query an LLM and then take action of what it says, it's disaster. That's why no one has agents. No one's automating anything. The use cases that are emerging, there are good ones, and they're very useful cases because it's very powerful technology, are not the use cases that a trillion dollar, the support Of trillion dollar company or a 500 billion dollar company like it is now. This is why there's also these shenanigans going on where, you know, the Nvidia deal with open AI where I'm going to pay you a hundred billion dollars for this and you're going to pay me A hundred billion dollars back. And then we're going to kind of hope that there's a stock market inflation. And in that arbitrage to somehow makes a difference. There was a great article about this in the journal recently. You know what the last time those type of deals were going on was? It was in the Enron era with all those fiber companies right before they all went out of business in the late 90s, early 2000s. So I think this is a mark, this shift towards a tension economy slop that we've seen in the last like month from Open AI is a signal that they're in trouble because it was super expensive To build these models and though they're cool they're not going to take over the economy and make them the only company left and be the last invention that anyone ever needs to make and Now they're like how do we make money we're making some like they do pretty well selling chat gpt subscriptions but they need to like 10x that to really capture their, their CapEx expenses. So they're now starting to flail. That's the way I see these two stories. So what we started with building a TikTok clone to what we ended with saying, uh, let's just try to make the chat feed as like whatever brainstem engaging we can. And by the way, they're going to sell ads on all this. Of course, they hired a major, a major ad expert onto their executive team earlier in the summer. I mean, this is what they're going to do is try to sell ads on just brainstem stimulation. I see this as the fall of like a once super ambitious, exciting company with visions for changing the world. And now they're hoping that just sort of like slop their way towards profitability and maybe other breakthroughs are coming and I'll be wrong. But then I'm not a stock market analyst and I'm just a technology guy. I'm just saying I think these two things together is not a good sign if you were really on the team of we're like a year away from open AI changing everything. So there we go. ([Time 1:28:09](https://share.snipd.com/snip/421a519e-42e4-490c-a498-2c4919529bff))